5

Edit: Because of the high number of downvotes on this question, I would appreciate if you could comment on why you think the question deserves so. I believe the question is clearly asking for the purpose of those vents. Suggestions and edits on how to make that clearer are much appreciated.

The servers are already stacked, and everything has been done as per the manufacturers instructions. I'm not asking if I'm doing it right, I'm asking why is that. Maybe the question belongs to another site on stack-exchange (I originally thought about the engineering site, since this is a engineering question, but assumed I would find the most expertise on this specific topic here).

We have a series of Dell rack servers and a few of them have vent holes on their top lid. This seems a bit odd since I would expect them to be designed for the purpose of being stacked. But if the units are meant to be stacked, the vent holes become useless.

I have found a similar question, Rack Mounting a Server with hot air vents on the top?, but the answers were focused on "do not worry about it". Well... then why are those hole vents there in the first place? Maybe something during production of the lids? Or to ease automatic assembly?

One of the comments in that question suggested reading the manufacturer's instruction. I found nothing on them.

Here's some pictures of the servers (well, pictures of the same model which I found on the internet):

  • Dell PowerEdge R715 PowerEdge R715 view from the top

  • Dell PowerEdge R820 - note that there's two rows of vent holes in the front as well. PowerEdge R820 view from the top

Community
  • 1
carandraug
  • 175
  • 5
  • Possible duplicate of [Rack Mounting a Server with hot air vents on the top?](http://serverfault.com/questions/396646/rack-mounting-a-server-with-hot-air-vents-on-the-top) – ewwhite Apr 04 '16 at 18:20
  • Then I'd suggest not worrying about it... – ewwhite Apr 04 '16 at 18:21
  • 2
    @ewwhite I referenced that same question on my own question. I specifically ask what is the purpose of those vents which is never addressed on the other question. – carandraug Apr 04 '16 at 18:36
  • But the point is that it doesn't matter why they are there. You know what the airflow for the server is (_front to rear_), and you know that the servers can also be safely stacked without gaps. Do you have a specific concern? – ewwhite Apr 05 '16 at 05:33
  • 1
    @ewwhite that is not the point, because that is not my question. I did not ask if there's an issue with blocking the vents, I asked why they are there. This is not a question of the type "can I do X?". And even if it was, I would expect to have some technical references to backup the "don't worry about it, I do it like that and never had issues" claims. – carandraug Apr 05 '16 at 10:58
  • Alternative explanation: Dell is beholden to the god of 1U servers, and the location of these vents is pleasing to Him. Do you want to anger the god of 1U servers? THEN WHY DO YOU QUESTION HIM? – Parthian Shot Apr 07 '16 at 15:41

3 Answers3

13

(Some people are hard to please...)

These are likely system exhaust vents and fit into the total design of the product. Please understand that the system is designed and optimized for rack mounting in a proper enclosure with separation between hot and cold air aisle. These servers are designed to be placed in adjacent rack units, and there's a small gap that fits within the envelope of the rack unit standard.

To allow space between adjacent rack-mounted components, a panel is 1⁄32 inch (0.031 inch or 0.79 mm) less in height than the full number of rack units would imply.

That's the primary use case for this type of server and drives the product development target. When I say "total design", that may include factors like thermal management, vibration, fan speed, power consumption and acoustics.

If you see vents and don't understand why they may be there, possibilities could include eliminating known hot-spots or being part of an internal air circulation strategy for different potential configurations. There's already internal baffling and shrouding to help direct airflow between zones. These may simply be exhaust vents for air, heat or both. Maybe the server vibrates less with their presence, or perhaps the vents change the timbre of the fan sound.

Sometimes, you may see a design element meant for an edge case, like a fully-loaded PCIe riser cage or for a minimal configuration like a single CPU and single power supply. Or maybe it's relief for a situation where the server is sitting and running on a workbench or table. Perhaps this type of venting allows the server to run with a lower fan speed and rely more on passive heat dissipation. That can address both acoustic and power efficiency goals.

For instance, the rear-right edge of my HP ProLiant DL360p Gen8 1U rack mount server is the location of the 10GbE NIC module and the embedded RAID controller. There's also room for a PCIe card in the same spot.

  • The present temperature at that specific location inside the server is 183°F (84°C).
  • The threshold for that location is 221°F (105°C).
  • There are perforations in the server's top cover at that location.
  • With a full PCIe load, these perforations serve as exhaust and it makes sense that there would be accommodations at that hotspot.

enter image description here

enter image description here

My suggestion is to use the server as designed and indicated by the manufacturer. If you don't have a full rack and proper datacenter environment or will be operating in less-than-ideal conditions, make sure that the server's vents are unobstructed. Monitor your system's vitals, since this equipment is designed to record, report and react to its surroundings. Otherwise, I wouldn't worry too much about this.

ewwhite
  • 197,159
  • 92
  • 443
  • 809
  • The references were a most interesting read. However, none ever mentions top vents. Your proposals boils down to temperature or sound but you use the words "likely" and "maybe", and again, with no references. Call me hard to please but coming from science I'm used to have such statements backed up with a reference. I have used a small stripe of paper and put it on top of those vents, in the small gap between servers. The stripe of paper didn't move. Maybe it's because the server is at low usage, the PCI slots are empty, or our sub-optimal separation between the cold and hot air aisles. – carandraug Apr 07 '16 at 13:48
  • 4
    You should contact Dell for a product-specific explanation. – ewwhite Apr 07 '16 at 14:09
  • 1
    @carandraug I'm not sure what reference you're after. As said, these servers are designed and built to placed into a rack. That's just....how it is, everywhere, the world over. It's what Dell expects you to do with them and they've built them accordingly. I doubt you're going to get a 100% proven explanation because they're not going to release their R&D documentation to you – Dan Apr 07 '16 at 14:23
  • 1
    @carandraug So, fine, take the science approach. Buy 4,000 of these servers in every possible configuration, rack them all up and examine the heat and airflow around those vent holes. Then buy 4,000 more, have someone fabricate 4,000 case tops without vent holes, set those up the same way you setup the first group and compare the two groups. Let us know how that turns out. – HopelessN00b Apr 07 '16 at 14:32
  • 1
    @carandraug You know, someone wrote you a well thought-out, extensive, detailed answer. Just because you're "coming from science" doesn't mean you need to be a rude asshole. "Thanks" would be nice. – ceejayoz Apr 07 '16 at 15:12
  • @ceejayoz I am not being an asshole. I upvoted the answer because it's a good answer. I'm just explaining why I'm not marking it as correct. I'm sorry if it came out as rude but believe a thank you is implied (I'm used to see the thank you and hellos edited out of questions). – carandraug Apr 07 '16 at 15:25
  • @ewwhite It's seems that my comment as come out as rude. I'm very sorry for that. I was honest saying that the references were interesting. I lol'ed at the cardboard boxes on top of the racks, and enjoyed the analysis of sounds. Thank you for the detailed answer, I learned a lot from it, it's quite detailed and I had upvoted it. My comment was explaining why I am not marking it as correct. I should have been more clear on the first comment but there's a character limit and I often see such pleasantries being edited out of questions. Thank you one more time. – carandraug Apr 07 '16 at 15:35
2

Those holes are IMHO there to allow the air to keep venting in the event a person fills the expansion slots. Cards in expansion slots don't have vent holes so each care is removing venting real estate.

Plus, if you are racking properly, the servers shouldn't actually touch each other. Granted the gap is small but it's enough to allow air to pass through those rear-ward, top vent holes.

Skeer
  • 110
  • 1
  • 8
  • Do you have any reference for that? I even went through some of Dell white papers on rack architecture and design but couldn't find a note about this. – carandraug Apr 04 '16 at 18:41
  • No I don't.. it's just logical to me. It'd be of zero benefit to the manufacturing process to have those holes. Punching them means the process took longer for each unit. Plus look at the number of holes on the rear of the case.. if you toss in 4 expansion card you;d removed a good 50% of the available ventilation holes. – Skeer Apr 04 '16 at 18:50
  • That does not explain: 1) purpose of the vents at the front of the top lid (maybe same reason as the expansion slots but for the 2.5 HD slots in the front); and 2) why not all of the rack servers (also with several expansion slots and very similar number of holes in the back), do not have the vents on top? – carandraug Apr 05 '16 at 11:56
1

Servers aren't really intended to be stacked directly on top of each other but for mounting in a 19'' cabinet. Typically there they don't quite touch and those vent holes won't be (completely) closed off.

For their actual purpose, I can only guess... Either to allow airflow and prevent excessive pressure differences when the main intake and outflow aren't quite balanced?

Or maybe to allow thermal circulation for when the server is switched off. Then you'd still want to have the possibility for circulation and a way for any remaining heat to escape.

HBruijn
  • 77,029
  • 24
  • 135
  • 201